Predicting a next alert in a pattern of alerts to identify a security incident
Disclosed herein is a system for predicting, given a pattern of triggered alerts, a next alert in order to identify malicious activity that is about to occur on resource(s) being monitored by a security operations center. A resource can include a server, a storage device, a user device (e.g., a personal computer, a tablet computer, a smartphone, etc.), a virtual machine, networking equipment, etc. Accordingly, the next alert is speculatively triggered in advance and a security analyst can be notified of a pattern of activity that is likely to be malicious. The security analyst can then investigate the pattern of triggered alerts and the speculatively triggered alert to determine whether steps to mitigate the malicious activity before it occurs should be taken.
1. A system comprising:
one or more processing units; and
a computer-readable storage medium having computer-executable instructions stored thereupon, which, when executed by the one or more processing units, cause the one or more processing units to:
learn a model based on a stored pattern of alerts, wherein each alert in the stored pattern of alerts is known to be indictive of a same security issue;
receive a plurality of alerts generated by an alert monitoring system deployed in association with a resource based on actions detected in association with the resource;
apply the model to the plurality of alerts received to determine that at least two alerts in the plurality of alerts received comprise at least part of the stored pattern of alerts;
predict, based on application of the model, a next alert that is to be generated in the stored pattern of alerts after the at least two alerts; and
present, in a graphical user interface, information associated with the next alert and the stored pattern of alerts.
2. The system of claim 1 , wherein the information includes a type of malicious activity likely to be associated with the next alert and the stored pattern of alerts.
3. The system of claim 1 , wherein:
the model determines a probability of the next alert being generated; and
the information includes the probability of the next alert being generated.
4. The system of claim 3 , wherein the probability of the next alert being generated increases as a number of alerts in the stored pattern of alerts increases.
5. The system of claim 3 , wherein the computer-executable instructions further cause the one or more processing units to:
compare the probability of the next alert being generated to a probability threshold;
determine that the probability of the next alert being generated is greater than the probability threshold; and
present, in the graphical user interface, the information associated with the next alert and the stored pattern of alerts based at least in part on determining that the probability of the next alert being generated is greater than the probability threshold.
6. The system of claim 1 , wherein:
the model determines an estimated time at which the next alert is to be generated; and
the information includes the estimated time at which the next alert is to be generated.
7. The system of claim 1 , wherein the computer-executable instructions further cause the one or more processing units to:
determine that a severity level associated with the next alert is greater than a severity level threshold; and
present, in the graphical user interface, the information associated with the next alert and the stored pattern of alerts based at least in part on determining that the severity level associated with the next alert is greater than the severity level threshold.
8. The system of claim 1 , wherein the computer-executable instructions further cause the one or more processing units to extract features from the individual alerts in the stored pattern of alerts, wherein:
the features include words used to describe individual alerts; and
the model is trained to map the words used to describe the individual alerts to a linguistic context and to predict the next alert based on identifying one or more words in the linguistic context that follow the words used to describe the individual alerts.
9. The system of claim 8 , wherein the model uses neural networks and natural language processing.
10. A method comprising:
learning, by one or more processing units, a model based on a stored pattern of alerts, wherein each alert in the stored pattern of alerts is known to be associated with a same security issue;
receiving a plurality of alerts generated by an alert monitoring system deployed in association with a resource based on actions detected in association with the resource;
applying the model to the plurality of alerts received to determine that at least two alerts in the plurality of alerts received comprise at least part of the stored pattern of alerts;
predicting, based on application of the model, a next alert that is to be generated in the stored pattern of alerts after the at least two alerts; and
presenting, in a graphical user interface, information associated with the next alert and the stored pattern of alerts.
11. The method of claim 10 , wherein the information includes a type of malicious activity likely to be associated with the next alert and the stored pattern of alerts.
12. The method of claim 10 , wherein:
the model determines a probability of the next alert being generated; and
the information includes the probability of the next alert being generated.
13. The method of claim 12 , wherein the probability of the next alert being generated increases as a number of alerts in the stored pattern of alerts increases.
14. The method of claim 12 , further comprising:
comparing the probability of the next alert being generated to a probability threshold;
determining that the probability of the next alert being generated is greater than the probability threshold; and
presenting, in the graphical user interface, the information associated with the next alert and the stored pattern of alerts based at least in part on determining that the probability of the next alert being generated is greater than the probability threshold.
15. The method of claim 10 , wherein:
the model determines an estimated time at which the next alert is to be generated; and
the information includes the estimated time at which the next alert is to be generated.
16. The method of claim 10 , further comprising:
determining that a severity level associated with the next alert is greater than a severity level threshold; and
presenting, in the graphical user interface, the information associated with the next alert and the stored pattern of alerts based at least in part on determining that the severity level associated with the next alert is greater than the severity level threshold.
17. The method of claim 10 , further comprising extracting features from the individual alerts in the stored pattern of alerts, wherein:
the features include words used to describe individual alerts; and
the model is trained to map the words used to describe the individual alerts to a linguistic context and to predict the next alert based on identifying one or more words in the linguistic context that follow the words used to describe the individual alerts.
18. The method of claim 17 , wherein the model uses neural networks and natural language processing.
19. A system comprising:
one or more processing units; and
a computer-readable storage medium having computer-executable instructions stored thereupon, which, when executed by the one or more processing units, cause the one or more processing units to:
receive a plurality of alerts associated with a plurality of resources being monitored by a security operations center, wherein an individual alert is generated by an alert monitoring system deployed in association with a particular resource or a particular group of resources based on one or more detected actions;
apply a model to identify, within the plurality of alerts, at least two alerts are in a stored pattern of alerts that indicates a security issue;
predict, based on application of the model, a next alert that is to be generated in the stored pattern of alerts after the at least two alerts; and
present, in a graphical user interface, information associated with the next alert and the pattern of alerts, wherein the information includes a probability of the next alert being generated.
20. The system of claim 19 , wherein the graphical user interface presents an option to provide feedback indicating whether the next alert and the stored pattern of alerts amount to a real security issue, thereby enabling supervised learning of the model.